High resolution gel electrophoresis was used to monitor the successive addition of dNMP residues onto the 3'-OH ends of discrete 5'-32P-primers, during DNA synthesis on natural templates. Resulting autoradiographic banding patterns revealed considerable variation in the relative rates of incorporation at different positions along the template. The pattern of "pause sites" along the template was unique for each of three different DNA polymerases (polymerase I (the "large fragment" form of Escherichia coli), T4 polymerase (encoded by bacteriophage T4), and AMV polymerase (DNA polymerase of avian myeloblastosis virus]. Most pause sites were not caused by attenuation of polymerization at regions of local secondary structure in the template. Assays of the accuracy of incorporation at different positions along the template (in which elongation was monitored in the presence of only 3 of the 4 2'-deoxynucleoside 5'-triphosphates) strongly suggested that the relative fidelity of DNA synthesis catalyzed by different polymerases depends on the position on the template at which the comparison is made. Primer-templates were constructed that permitted comparison of elongation during synthesis on a single-stranded template with that during polymerization through a double-stranded region (wherein elongation required concomitant displacement of a strand annealed adjacent to the 5'-32P-primer). Although strand displacement DNA synthesis catalyzed by polymerase I occurred approximately ten times more slowly than synthesis in the same region of a single-stranded viral template, most of the pause sites were the same in the presence or absence of "tandem" primer. Electrophoretic assays of the fidelity of DNA synthesis suggested that an increased tendency toward misincorporational "hotspots" occurred when elongation required concomitant strand displacement.